All nodes must have a watchdog, as we can run into a kernel panic that crashes the node and becomes unresponsive.